PTN Clock Synchronization Technology and Its Applications

    As operators seek to replace their conventional Time Division Multiplexing (TDM) transport networks with Packet Transport Networks (PTN), clock synchronization comes into consideration. There are two requirements for PTN synchronization:

    (1) It must be able to bear TDM services and provide a clock recovery mechanism.

    (2) PTN should have a high-precision network reference clock for synchronization of network nodes.

1 Overview of Synchronization Technologies
Clock synchronization includes frequency synchronization and time synchronization. The former implies identical timing intervals, while the latter implies identical starting time as well as intervals. Different wireless standards have different requirements for clock bearing.
Since 2004, the ITU-T Q13/SG15 has made a number of proposals relating to PTN, including G.8261 (defining general requirements), G.8262 (defining equipment clock performance), and G.8264 (defining system architectures and synchronization function modules).
IEEE 1588 standards were released in 2002 with the view of defining Precise Time Protocol (PTP) for a LAN multicast environment.

    However, standards released specifically for this purpose could not be applied to telecommunications environments with greater complexity. IEEE 1588v2 was therefore issued in 2008 for application in telecom networks[1-5]. The IETF Network Time Protocol (NTP) implements time synchronization between users and between time servers in the Internet.

2 Synchronous Ethernet Technology
Physical layer synchronization technology is widely applied in traditional Synchronous Digital Hierarchy (SDH) networks. Each node in an SDH can extract the clock from the physical links or obtain the clock from external synchronous interfaces. A node selects the best quality clock from among the multiple clock sources, and sets local clocks accordingly. Then the locked-in clock is transported to equipment at lower layers. Each layer of the network becomes synchronized to the Primary Reference Clock (PRC) through level-by-level locking. Similar physical-layer synchronization technology can be used in packet networks, as shown in Figure 1.

2.1 Principle of Synchronous Ethernet
Synchronous Ethernet technology in PTN uses Ethernet link steams to recover clocks. Ethernet physical layer coding adopts 4B/5B (FE) and 8B/10B (GE) technology, in which one additional bit is inserted into every four bits. In this way, four consecutive 1s or four consecutive 0s will not appear in any transported data stream, and the clock information is effectively contained. High-precision clocks are used to transport data from the Ethernet source-end interface, and are recovered and extracted at the receiving end.

    The principle of synchronous Ethernet is shown in Figure 2. A high-precision clock is embedded into the physical layer chip of equipment at the source side (Node A). This clock is used by the chip to send data out. The physical layer chip of equipment at the receiving side (Node B) extracts this clock from data streams. Precision is not damaged in this process and synchronicity with the source end is maintained. The principle of clock transport in synchronous Ethernet is similar to that of an SDH network—the clock is recovered from the Ethernet physical link. As a result, the quality of recovered clock is not influenced by link service traffic, and clock tree deployment and quality is the same as that of SDH/SONET. This satisfies the timing interface index of G.823.  

2.2 SSM Algorithm for Synchronous Ethernet
Synchronization Status Message (SSM) algorithm originates from SDH clock synchronization control. Its application rules and clock selection algorithm conform to the ITU-T G.781 standard, and control of synchronous Ethernet retains SDH network characteristics. Moreover, based on the conventional clock network, SSM supports synchronous Ethernet by building in Ethernet Synchronous Message Channel (ESMC). As described in G.8264, ESMC is a unidirectional broadcast protocol channel of the Media Access Control (MAC) layer used for transporting SSMs between different equipment. The equipment selects the best clock source based on SSM in ESMC messages.

    Standard SSM algorithm is effective for network clock synchronization. However, it has two weaknesses:

    (1) It cannot satisfactorily handle the problem of synchronous clock ring.

    (2) It has clock signal attenuation. With an increase of synchronous links, drift caused by noise and temperature changes in synchronization allocation deteriorate timing reference signals. Consequently, the number of synchronized Network Elements (NEs) is limited on the same synchronous link, and standard SSM has difficulty tracing and counting nodes.

    For PTN, ZTE proposes an improved and extended SSM algorithm. ESMC packet uses two Type-Lengh-Values (TLVs) to transport SSM messages. Following the ITU-T standards, the first TLV transports original SSM byte information at the synchronization quality level. The second TLV is used for path protection. The improved algorithm has the following advantages:

• It fundamentally prevents clock ring;
• It automatically selects the optimal (shortest) route in the event of multiple clock paths;
• If there is a route to the master clock, NEs will trace the master clock instead of entering  a state of free oscillation;
• The algorithm is ideal for processing because it supports distributed processing at lower layers and each NE is in an equal position;
• Due to the direct use of standard S1 bytes, there is no problem interworking with other vendors’ equipment.

3.1 Network Time Protocol (NTP)
Before the advent of IEEE 1588v2, there were three main types of PTN time synchronization protocol: time protocol, daytime protocol and NTP. NTP is fulfilled purely by  low-precision software; the widely used NTPv3, for example, achieves synchronization precision of 10 ms. IETF is now proceeding with NTPv4 standards that support IPv6 and dynamic discovery servers. Synchronization precision of NTPv4 is expected to reach a 10 μs level. However, the stability and precision of NTP still fails to meet the high requirements of telecom networks.

3.2 IEEE 1588v2 Protocol

3.2.1 Implementation Principle of IEEE 1588v2
IEEE 1588v2 unifies time and frequency synchronization. In the future, it will be suitable for inter-office time-frequency transport of different platforms. It is capable of unidirectional frequency transport using Time over Packet (ToP) based on IEEE 1588v2 timestamp. It also realizes time synchronization by using IEEE 1588v2 protocols. Therefore, it is widely applied in PTNs.

    IEEE 1588v2 time synchronization is achieved by encoding time signals through use of master and slave clocks. Both master and slave time are synchronized by bidirectional message interaction as well as by employing network symmetry and delay measurement technology.
According to the 1588v2 principle shown in Figure 4, Delay= (T2-T1+T4-T3)/2; Offset= (T2-T1-T4+T3)/2.

    Messages such as Sync, Follow_Up, Delay_Req, and Delay_Resp are delivered between the master and slave clocks. Referring to the four values T1, T2, T3 and T4, the delay and offset between the master and slave can be determined.

    Synchronization messages include general messages and event messages. A general message, such as Follow_Up, does not handle timestamp itself; it can carry accurate
delivering/receiving time of event messages such as Sync. In addition, it implements network configuration and management, and fulfills communication among PTP nodes. An event message needs to deal with timestamp, and can carry timestamp or not. Depending on the event message timestamp or general message timestamp, the slave clock can determine the path delay and offset between the master and slave clocks.  

3.2.2 Clock Types
Based on protocols such as Ethernet, IPv4/IPv6, and User Datagram Protocol (UDP), IEEE 1588v2 defines three basic clock types: Ordinary Clock (OC), Boundary Clock (BC), and Transparent Clock (TC).

 
    OC is a single-port device acting as either the master or slave clock. Only one master clock is permitted in a synchronous domain. Frequency, accuracy, and stability of the master clock is directly related to the performance of the entire synchronous network. PRC or GPS is generally adopted for network synchronization. Slave clock performance determines the timestamp precision and sync message rate.  

    
    As a multi-port device, BC can connect with many OCs/TCs. Among the multiple ports of BC, one serves as the slave port, connecting to the master port of the master clock or other BCs. The other ports serve as master ports, connecting to the slave port of the slave clock or BC at the next layer. They can also serve as backup ports. 

    TC is used for connecting the master clock with the slave clock. It forwards interacted synchronous messages between the master and slave clocks in a transparent way. It also calculates local buffering treatment time for synchronous messages—such as Sync and Delay_Req—and writes this time into the CorrectionField byte block of the sync messages. The slave clock calculates delay and offset based on byte values and sync message timestamp in order to realize synchronization. TC can be divided into Edge-to-Edge (E2E) TC, and Peer-to-Peer (P2P)TC.  

3.2.3 Delay in IEEE 1588v2
Delay is a major factor affecting precision of an IEEE 1588v2 system.

    (1) Timestamp Treatment Delay 
    IEEE 1588v2 timestamp treatment is performed by hardware. The timestamp processor is located between the physical layer and MAC layer, as shown in Figure 5.

 
    Hardware timestamp treatment compensates for the time taken by IEEE 1588v2 protocol frame to pass through a protocol stack. Therefore, precision of message delivery and timestamp receiving at the port is guaranteed.

    (2) Node Buffering and Path Delay
    IEEE 1588v2 defines E2E TC and P2P TC for compensating node buffering delay. There are two methods of transport path compensation: delay request-response, and peer delay.
Delay request-response works with E2E TC. This TC marks timestamp treatment in the packet at ingress and egress, while the slave completes all the calculations on time delay compensation. 

    Peer delay works with P2P TC. This TC calculates time delay among endpoints, and each endpoint interacts with the TC respectively to compute P2P time delay. Based on these calculations, the slave then calculates delay compensation. 

3.2.4 Fulfillment of IEEE 1588v2 in PTN
IEEE 1588v2 synchronization precision is affected by many factors in real network deployment. Its application in complex network environments, such as hybrids of microwave and switching networks, is still in the research stage. IEEE 1588v2 can achieve precision of 100 ns in a pure PTN testing network. However, due to the complexity of network time delay and lack of control over IEEE 1588v2 bidirectional path asymmetry, there are unpredictable risks when relying solely on IEEE 1588v2 protocol and analytic algorithm to adapt to real networks. When a network is heavily loaded, frequently sent IEEE 1588v2 message packets are easily influenced by service packets, and this greatly impacts time delay precision. However, by decreasing the frequency of sending message packets, the time convergence rate may be slowed. In addition, the IEEE 1588v2 algorithm must be compensated with bidirectional path asymmetry—achieved by optical fiber—in actual network construction. Timing error of optical fiber asymmetry is first measured by costly time synchronization testers and oscilloscopes, and is followed by asymmetrical time delay compensation. Owing to the large number of PTN nodes, compensation work requires many engineers. Instruments such as time synchronization testers and oscilloscopes are also cumbersome. It is therefore difficult to apply IEEE 1588v2 widely in real network construction, and opinions still vary about its feasibility.

    ZTE Corporation proposes a synchronous Ethernet-based IEEE 1588v2 time transport solution to the problem. The core idea is to establish a highly controllable network where clocks are isolated from time, eliminating unpredictable risk. This is beneficial for fast convergence of time synchronization in order to implement IEEE 1588v2 based on steady frequency synchronization of the synchronous Ethernet physical layer.

    Moreover, it reduces the frequency of IEEE 1588v2 message packet delivery. Time precision is not affected even if the network is heavily loaded, which improves reliability and precision of PTN time synchronization. To solve the engineering problem in PTN asymmetry measurement, ZTE configures its access-layer PTN equipment with a time error measurement function that quickly, easily, and accurately performs measurements without the need for professional instruments. 

4 Typical Applications

4.1 Application of Synchronous Ethernet
Similar to SDH, synchronous Ethernet supports both ring networking and tree networking. Generally, RNC provides the clock source, and clock information arrives at every base station via synchronous Ethernet in order to maintain synchronicity throughout the entire network. Tree networking does not support clock routing protection. However, in ring networking, when current clock routing fails, related NEs can receive messages such as warning and SSM, and trace the source clock from other directions in order to protect clock routing. Figure 6 shows examples of synchronous Ethernet networking. 

   
    Jittering of synchronous messages is intensified after they have been transported by NEs. Therefore, optimal clock quality is achieved when network equipment is able to trace the clock source using the shortest possible path. ZTE’s PTN adopts an extended SSM algorithm which includes the number of nodes the clock has passed through. In any case, NEs are able to trace the clock source using the shortest possible path.

    NE C can use point B or D to trace clock information with its source at A. The clock passes through only one node when tracing from point B, and two nodes when tracing from point D. To obtain higher clock quality, ZTE’s PTN automatically chooses the clock traced from B.

4.2 Applications of IEEE 1588v2

4.2.1 Replacing GPS at Base Stations
Typical IEEE 1588v2 application involves replacing GPS at Radio Access Network (RAN) base stations. Installing a GPS antenna at a TD-SCDMA or CDMA2000 base station requires an area of 120 degrees to be cleared, and this can be difficult given environmental constraints. GPS is difficult to install indoors and underground. Moreover, GPS is expensive and has a high failure rate. If PTN can provide time synchronization for base stations, it can replace GPS (or act as the GPS backup) and offer higher security guarantee for RAN.

    Figure 8 shows an example of IEEE 1588v2 replacing GPS at base stations. Only one NE in the PTN needs to receive time messages from GPS; for example, through the 1PPS+ToD interface. PTN delivers the time messages to other NEs by IEEE 1588v2 protocol, then the messages arrive at the base stations via Ethernet or other interfaces. Time synchronization among all the base stations is fulfilled accordingly.

    Base stations support either IEEE 1588v2 protocol or time interfaces. If supporting IEEE 1588v2 protocol, PTN works in TC mode; if not, PTN works in BC mode.

4.2.2 Frequency Recovery
IEEE 1588v2 is also used to recover frequency by ToP. Most operator networks are ordinary data networks, not supporting synchronous Ethernet. In this case, IEEE 1588v2 can be used to obtain time frequency in these ordinary networks.

    Figure 9 shows an example of frequency recovery 1588v2 networking. When the network between equipment A and equipment B is an ordinary data network, IEEE 1588v2 Sync messages are transported from A to B via the ordinary data network. Equipment B uses IEEE 1588v2 to recover the clock. The recovered clock is taken as the reference source of B, with which the service clock is recovered.

5 Conclusions
Research into PTN clock synchronization technology is deepening. ZTE proposes an extended synchronous Ethernet SSM algorithm and IEEE 1588v2 solution based on synchronous Ethernet. This solution plays an important role in raising the precision of time synchronization in PTN and reducing the level of engineering complexity. PTN clock synchronization technology has extensive application possibilities, especially in RAN, TDM services, M2M real-time data acquisition, and VIP private networks.

References
[1] ITU-T G.8261.Timing and synchronization aspects in packet networks [S]. 2006.
[2] ITU-T G.8262.Timing characteristics of synchronous Ethernet equipment slave clock (EEC) [S]. 2007.
[3] ITU-T G.8264. Distribution of timing through packet networks [S]. 2008.
[4] IEEE 1588.IEEE standard for a precision clock synchronization protocol for networked measurement and control systems [S]. 2008.
[5] MEF 22. Mobile backhaul implementation agreement phase1 [S]. 2009.

3 Time Synchronization Technology
Time synchronization technology is a progression from frequency synchronization. By using the packet protocol data unit to carry clock/time messages, PTN time synchronization technology can synchronize the master clock and slave clock. Figure 3 shows the basic principle.

[Abstract] Clock synchronization is an important issue for packet transport networking. Current clock synchronization technologies include synchronous Ethernet, IEEE 1588v2, and Network Time Protocol (NTP). However, individually, these technologies are beset with certain problems. Synchronization Status Message (SSM) algorithm for synchronous Ethernet standards suffers clock ring, and has difficulty tracing and counting nodes. An extended SSM algorithm can improve clock synchronization. NTP is too imprecise to meet the requirements of telecom networks, yet IEEE 1588v2 alone can lead to slow convergence time and influence time delay precision when the network is heavily loaded. ZTE therefore proposes an IEEE 1588v2 solution based on synchronous Ethernet in order to effectively raise the precision of Packet Transport Network (PTN) time synchronization.